Abstract: Aiming to solving the limitations of existing lightweight image super-resolution reconstruction algorithms in terms of feature richness and accuracy, proposes a lightweight image super-resolution reconstruction algorithm based on multi-path feature calibration. To achieve multi-scale global to local feature extraction, multi-field feature extraction blocks are first constructed before deep feature extraction, and feature calibration is coordinated in channel and spatial dimension to enhance the global understanding of the network’s overall structure. Then, to obtain more comprehensive and information-rich image features and enhance the representation ability of the network, a multi-channel feature calibration module is proposed, which can extract the feature information of different granularity from different branches of the multiple channels. Finally, a lightweight image super-resolution reconstruction network based on multi-path feature calibration is constructed by modelling the contextual dependencies from the multi-scale space and channel dimensions to fully exploit the spatial information and channel features of the image, and to design the multi-field spatial attention and the channel calibration attention, respectively. Large numbers of experimental results show that the proposed algorithm achieves a high balance between parameters and performance, especially for images with complex structures and rich texture details. Especially on the Urban100 dataset with complex texture, the proposed algorithm improves the PSNR index by at least 0.1 dB compared with the compared lightweight methods.